https://www.irjet.net/archives/V5/i11/IRJET-V5I11270.pdf

1. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 05 Issue: 11 | Nov 2018 www.irjet.net p-ISSN: 2395-0072
© 2018, IRJET | Impact Factor value: 7.211 | ISO 9001:2008 Certified Journal | Page 1391
COLOR BALANCE AND FUSION FOR UNDERWATER IMAGE
ENHANCEMENT: SURVEY
JENI MONI1, ANJU J PRAKASH2
1M. Tech. Student, Computer Science and Engineering, Sree Buddha College of Engineering, Kerala, India.
2Assistant Professor, Computer Science and Engineering, Sree Buddha College of Engineering, Kerala, India.
---------------------------------------------------------------------***----------------------------------------------------------------------
Abstract - Underwater images suffer poor visibility mainly
due to scattering and absorption effects. There have been
several approaches to restore and enhance the visibility of
degraded underwater images. Singleimageapproachwithout
the need of specialized hardware and knowledge about
underwater conditions or scene structure is a novel approach
for underwater image enhancement. Color compensated and
white balanced version of original image as inputs as well as
their associated weight maps undergoes a multi-scale fusion
strategy. Thus enhanced images are characterized by better
exposedness of the darkregions, improvedglobalcontrast, and
edges sharpness.
Key Words: Underwater, Image fusion, White-balancing.
1. INTRODUCTION
Underwater imaging is an important area in research and
present technology. There are several rare attractions in
underwater environment such as amazing landscapes,
marine animals and mysterious shipwrecks. Scattering and
absorption of light are the major reasons for low contrast
and low clarity of underwater images. Absorption
considerably reduces lightenergyanditdependsuponmany
factors such as salinity and turbidity of water, amount of
suspended particles etc. Light scattering causesdeflection of
the ray from a straight path due to irregularities in the
propagation medium, particles etc. They results in foggy
appearance, low contrast and fading of colors. Also image
captured in underwater is hazy due to several effects of
underwater medium. These effects are caused by the
suspended particles in underwater. Water absorbs light
wavelength to different degrees. Longer wavelengths get
absorb in water first and shorter wavelength appear at a
long distance. Water depth is highly correlated with color
perception. The penetration of the visible spectrum colors
depends on the depth of the water and wavelength.
Disappearance of color in underwater occurs in the same
order as they appear in the color spectrum and therefore it
results in bluish tone of underwater images.
There are many strategies and methods for enhancing and
restoring underwater images. Traditional enhancing
techniques such as histogram equalization and gamma
correction show strong limitations. It is also possible to
enhance images using specialized hardware, wavelength
compensation, wavelet strategy and dark channel dehazing.
These all strategies can enhance images but not much
efficient for practicability due to some limitations.
Proposed method is an effective approach which is able to
remove the haze and enhance image based on a singleimage
captured with a conventional camera. It builds on the fusing
of two images that are directly derived from color
compensated and white-balanced version of the original
degraded image. The white balancing stage removes
undesired color cast induced by underwater light scattering
and produce natural appearance of underwater images. It
reduces the quantization artifacts introduced by domain
stretching. A well-known white balancing method Gray-
World algorithm is used which can achieves good visual
performance for reasonably distorted underwater images.
The reddish appearance of high intensity regions in the
image is also well corrected since the red channel is better
compensated.
Multi-scale implementation of fusion is an effective fusion
based approach, relying on gamma correction and
sharpening to deal with the hazy nature of the white
balanced image. The weight maps suchasLaplaciancontrast
weight, saliency weight, saturation weight maps are used
during blending in such a way that pixel with a high weight
value are more represented in the final output image. It also
can enhance the quality of the underwater images. The
enhanced image after applying proposed method is given in
Fig -1.
Fig -1: Degraded and enhanced underwater image
2. LITERATURE SURVEY
Various methodsare used forenhancing underwaterimages.
Some of them are discussed below.
2.1 Underwater Image Restoration Based on
Image Blurriness and Light Absorption
Yan-Tsung Peng et al. [2] proposed an accurate depth
estimation methodforrestoringunderwaterimagesbasedon
image blurriness and light absorption. It can be used in the
image formation model to enhance and restore the degraded

2. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 05 Issue: 11 | Nov 2018 www.irjet.net p-ISSN: 2395-0072
© 2018, IRJET | Impact Factor value: 7.211 | ISO 9001:2008 Certified Journal | Page 1392
underwater image. It is possible to restore underwater
images properly because scene depth is not estimated via
color channels. The proposed method is provided with more
accurate BL and depth estimation. First, BL is selected from
blurry regions in an underwater image. Then, based on the
BL, the depth map and the TMs are achieved to restore scene
radiance. The flowchart of the proposed method is shown in
Fig -2.
Fig -2: Flowchart of proposed method
Blurriness is an important measure of depth. Depthisnot
estimated by using image blurriness alone, both image
blurriness and light absorption are also considered.
Blurriness BL is determined from candidate BLs estimated
from blurry regions. The most comprehensive underwater
image restoration techniques are then used. By considering
BL, depth estimation based on light absorption can handle
artificial lighting. Water absorbs more light as the light rays
travel throughlonger distanceinthewater. Artificiallighting
is occasionally used to provide sufficient light for taking
underwater images and videos. Artificial lighting in an
underwater image forms a bright foreground. The light
arising from an artificial lighting source is reflected by
foreground objects. It travels less far in the water and is less
absorbed and scattered. Artificially illuminated bright
foreground pixels is less improved by a restoration method
than background pixels. If the BL of an underwater image
with dim artificial lighting, the restoration using the depth
map derived by the red channel map would regard those
bright pixels as being close and not over-compensate their
color. When BL is bright, the red light from the background
pixels would attenuate more than that of the foreground
pixels, which could be correctly interpreted as scene depth.
The proposed method can create better restoration and
enhancement results in different underwatercolortonesand
lighting conditions compared to other underwater image
restoration methods.
2.2 Low Complexity Underwater Image
Enhancement Based on Dark Channel Prior
Hung-Yu Yang et al. [3] proposed low complex and
efficient underwater image enhancement method based on
dark channel prior. This approach consists of two main
procedures. First, estimation of airlight by calculating dark
channel prior and depth map is generated by using median
filter. Second, to further enhance the visual quality of
underwater image, an unsupervisedcolorcorrectionmethod
is used to improve the color contrast of the object. The flow
diagram of the proposed method is shown in Fig -3.
Fig -3: Flow diagram of proposed method
The atmospheric light is estimated by using dark
channel prior. The low intensities in the dark channel are
mainly due to the factors like shadows, colorful objects, dark
objects etc. In the dark channelpriormethod,thesoftmatting
algorithm is employed to eliminate the block effect of the
transmission and to reconstruct a better image. It requires
heavy computing resources and several iterations for
smoothing and optimizing the transmission.Inordertosolve
the problem, median filter is employed for the observed
image directly to obtain the smoothed transmission. The top
brightest pixels in the darkchannelarepickedoutandamong
these pixels; the pixels withhighestintensitiesareselectedas
the atmospheric light. Then an efficient color correction
method is used. Since underwater images have high blue
color when compared with remaining colors, the blue color
can be used to increase the green and red colors for making
the image balanced. The highest blue color is set as a target
mean and the remaining color channels are determined with
a multiplier to obtain a color balanced image.
2.3 Underwater Image Enhancement by Wavelet
Fusion
Amjad Khan et al. [4] proposed a wavelet-based fusion
method to enhance the hazy underwater images by
addressing the low contrast and color alternate problems.
Initially, the hazy degraded underwater image is replicated
into two categories. These categories are processed in
parallel to improve the image contrast and quality. The
wavelet based fusion process consists of a series of high pass
and low pass filter banks.
Contrast limited adaptive histogram equalization is a
form of adaptive histogram equalization. It is adopted for
enhancing the contrast and quality of underwater image by
clipping theunnecessaryregionfromthehistogram.Thelimit
for clipping is defined by the normalization of the histogram

3. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 05 Issue: 11 | Nov 2018 www.irjet.net p-ISSN: 2395-0072
© 2018, IRJET | Impact Factor value: 7.211 | ISO 9001:2008 Certified Journal | Page 1393
and size of the neighborhoodregionswithinthepixeldomain.
The region which is clipped is not discarded but equally
redistributed among all histogram bins. The contrast limited
adaptive histogram equalization is applied to all the three
color channels to improve the contrast of all tints availablein
the image. The flowchart of proposed method is shown in
Fig -4.
Fig- 4: Flowchart of proposed method
A sequence of low pass and high pass filter banks are
used to eliminateunwantedlowandhighfrequenciespresent
in the image and make the fusion process convenient. Two
levels of decomposition are employed here. There are two
steps in level one; the first step is obtained by applying the
low pass and high pass filters along with down-sampling on
the rows of the input image. Thus horizontal approximations
and horizontal details are generated. Then the columns in
horizontal coefficientsare filtered and down-sampled.Atthe
secondlevel of decomposition, thedecomposedapproximate
image of the level one becomes the input image. The process
is repeated to scale down coefficients. Each input image are
decomposed into wavelet coefficients by using the same
procedure. Finally, the both decompositions are fused by
using coefficients of maximum values.
2.4 Underwater Image Enhancement Using
Wavelength Compensation and Dehazing
John Y. Chiang et al. [5] proposed an approachtorestore
underwater images thatcombinesa dehazingalgorithm with
wavelength compensation.Thedistance betweenthecamera
and the object was estimated using dark-channel prior, the
haze effects were removed by the dehazing algorithm. Then
the scene depth is estimated from the residual energy ratios
of each wavelength in the background light of the image.
Amount of suspended particles and salt ratio varies with
time, location, and season. Reverse compensation is also
conducted according to the amount of attenuation of each
wavelength to restore the image.
Haze increases with distance in a hazy environment. By
evaluating the concentration of haze in a single image, it is
possible to predict the distance between the object in the
scene and camera. The calculation based on dark channel
prior is based on blocks and thus creates a less accurate
depth map. Image matting is also applied to repartition the
depth map and it requires input of a preliminary partitioned
depth map and the original image. Objects are identified by
using the relationship between the mean color and the
covariance of a local area within the image. Preliminary
partitioned depth map is then corrected by using the
relationship of objects themselves. Next, the estimation of
the distance between theobjectsandwatersurfaceiscarried
out to correct the color cast induced by water depth. Depth
estimation for each point in the image is needed to attain
corresponding energy compensation at the various depths.
Artificial light source are often supplemented during image
capturing process in order to avoid insufficient lighting
problems. Therefore it is necessary to detect luminance
contributed by an artificial light source avoid over-
compensation. Making accurate estimation of the rate of
energy attenuation is not possible because it varies with
season, time etc.
2.5 Underwater Image Enhancement by Dehazing
WithMinimumInformationLossandHistogram
Distribution Prior
Chongyi Li et al. [6] proposed a systematic method for
single underwater image enhancement whichcanyieldstwo
categories of enhanced output. The proposed method
consists of mainly two parts: underwater image dehazing
and underwater image contrast enhancement. An efficient
dehazing algorithm based on a minimum information loss
principle and optical properties of underwater imaging is
used. It can restore the visibility, color and natural
appearance of degraded underwater images. An effective
and simple contrastenhancementalgorithmisalsoproposed
based on a kind of histogram distribution prior, which
enhances brightness and contrast of underwater image.
There are two categories of enhanced images. First category
with natural appearance and relatively original color is
suitable for display. The second category with high
brightness and contrast can be used for extracting more
valuable information and unveiling more details. These two
categories of enhanced images can be used for several
applications.
Underwater image dehazing algorithm is composed of
three main processing steps: global background light
estimation, medium transmission map estimation, and
adaptive exposure map estimation.
1) Global background light estimation
It is often estimated as the brightest color in an underwater
image. A hierarchical stretching technique based on quad-
tree subdivision is used for estimating global background
light. The effects of suspended particles are removed by
using dark channel prior method. Finally the global
background light is estimated according to the properties of
light travelling in the water after removing the disturbances
of bright objects.
2) Medium transmission map estimation

4. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 05 Issue: 11 | Nov 2018 www.irjet.net p-ISSN: 2395-0072
© 2018, IRJET | Impact Factor value: 7.211 | ISO 9001:2008 Certified Journal | Page 1394
The enhanced image dependsontheselectionofthemedium
transmission map. A medium transmissionfunctionmapsan
input pixel value to an output value. An optimal medium
transmission map is selected to reduce loss of information.
In order to minimize information loss, the medium
transmission map of the most degraded channel is
estimated. Estimation of the medium transmission map of
red channel is more suitableforrestoringunderwaterimage.
3) Adaptive exposure map estimation:
Underwater image dehazing algorithm may cause dark and
bright regions become too dark or too bright. In order to
avoid this, it is used for better visual quality of underwater
image. Adaptive exposure map can balance too dark and too
bright regions.
Underwater image contrast enhancement is based on
the statistics of histogram distribution of natural-scene
images. The histogram distributionsofnatural-sceneimages
are more consistent and wider. This inspired to adjust the
histogram distributions of underwater images to attain
better contrast and brightness. Approximate histogram
matching is used to adjust the histogram distributions of
image obtained by underwater dehazing algorithm.
2.6 Underwater Image Enhancement Based On
Contrast Adjustment via Differential Evolution
Algorithm
Emre Gur Guraksin et al. [7] proposed an underwater
image enhancement using differential evolution algorithm.
Initially, the underwater image has been separatedintoRGB
color components and contrast of each component is
enhanced. After the contrast enhancement procedure,
concatenation procedure is done to the red, green and blue
components in order to achieve the colored image. In order
to determine the limit of contrasts, differential evolution
algorithm is employed while improvingthecontrastsofeach
component. Finally, by using unsharp masking principle,
underwater image is sharpened.
Histogram of an image is a measure of the frequency of
all color in the image. Insufficient lighting produces low
contrasted images. Contrast stretching can be used for
improving the contrast of these types of images. The main
idea behind the contrast stretching is to improve the
dynamic range of the gray levels in the processing image.
Differential evolution algorithm is a powerful and simple
population based stochastic search technique for resolving
global optimization problems. Mutation, crossover and
selection are the three main operators used in differential
evolution algorithm. In mutation operation, a mutant vector
is created for each target vector in the present population.
The mutation vector is created by using three randomly
chosen vectors with a parameter formutation.Aftercreating
mutant vector, it belongs to crossover operation with target
individual which creates a trial solution. After the crossover
operation, a selection process is employed to improve the
solution. The selection process in this algorithm is distinct
from the other evolutionary algorithms.
3. CONCLUSION
The literature survey could find a number of existing
underwater image enhancementmethods.Traditional image
enhancement and restoration methods cannot compensate
the contrast degradation of underwater images. Also it has
strong limitations for practical applications. In order to
overcome these limitations and issues, the proposed
framework builds on the multi-scale fusion principle and
does not require additional information than the single
original image. This will be able to enhance a wide range of
underwater images with highaccuracy, beingableto recover
important faded features and edges.
REFERENCES
[1] C. O. Ancuti, C. Ancuti, C. De Vleeschouwer and P.
Bekaert, "Color Balance and Fusion for Underwater
Image Enhancement," in IEEE Transactions on Image
Processing, vol. 27, no. 1, pp. 379-393, Jan. 2018.
[2] Yan-Tsung Peng and Pamela C, “Underwater Image
Restoration Based on Image Blurriness and Light
Absorption,” IEEE transactions onimageprocessing,vol.
26, no. 4, April 2017
.
[3] H. Yang, P. Chen, C. Huang, Y. Zhuang and Y. Shiau, "Low
Complexity Underwater Image Enhancement Based on
Dark Channel Prior," 2011 Second International
Conference on Innovations in Bio-inspired Computing
and Applications, Shenzhan, 2011, pp. 17-20.
[4] A. Khan, S. S. A. Ali, A. S. Malik, A. Anwer and F.
Meriaudeau, "Underwater Image Enhancement by
Wavelet Based Fusion," 2016 IEEE International
Conference on Underwater System Technology: Theory
and Applications (USYS), Penang, 2016, pp. 83-88.
[5] J. Y. Chiang and Y. Chen, "Underwater Image
Enhancement by Wavelength Compensation and
Dehazing," in IEEE Transactions on Image Processing,
vol. 21, no. 4, pp. 1756-1769, April 2012.
[6] C. Li, J. Guo, R. Cong, Y. Pang and B. Wang, "Underwater
Image Enhancement by Dehazing with Minimum
Information Loss and Histogram Distribution Prior," in
IEEE Transactions on Image Processing, vol. 25, no. 12,
pp. 5664-5677, Dec. 2016.
[7] G. E. Guraksin, U. Kose and O. Deperlıoglu, "Underwater
Image Enhancement Based on Contrast Adjustment via
Differential Evolution Algorithm," 2016 International
Symposium on Innovations in Intelligent Systems and
Applications (INISTA), Sinaia, 2016, pp. 1-5.